Associations Between Firearm and Suicide Rates: A Replication of Kleck (2021)

Abstract

Objective

Using data from n = 194 nation-states, Kleck found that firearm availability was only associated with firearm suicide rates, but not total or non-firearm suicides. He thus concluded that while firearm availability influences how people commit suicide, it does not affect total numbers. However, the study contains numerous logical and methodological issues and is at odds with the evidence base. Therefore, I attempt to reproduce the original results.

Method

I reproduce the original study’s methods: ordinary least squares regression, weighted by the square root of the population, with log-transformed suicide rates and three separate firearm availability measures: global estimates from the Small Arms Survey, proportion of suicides committed with firearms, and a European Union survey of firearm ownership. I also test several methodological variations and include U.S. suicide data.

Results

In contrast to Kleck, global analyses with Small Arms Survey data found a significant and positive association between firearm availability and total suicides, as did U.S. analyses. Analyses with other firearm availability measures comported with the original study, finding no association.

Conclusion

The main result in Kleck failed to reproduce, finding instead a significant association between firearm availability and suicide rates, as did U.S. analyses. While reproductions of Kleck’s other analyses continued to show no association, they were based on unreliable methods. I therefore reject Kleck’s conclusion that that firearm availability does not influence suicide rates.

Highlights

• Using data global data, I find firearm availability is positively associated with suicide rates.

• I identify serious flaws in the logic and methods of Kleck and an earlier review.

• For transparency, data and code have been archived on a public repository.

Keywords:

• Cross-national
• firearms
• open science
• replication

DATA AVAILABILITY STATEMENT

All data and analytical code have been archived on a public repository: https://doi.org/10.26180/14600337.

Notes

1 One of the analyses does find an association between firearm availability and non-firearm suicide rates, though it is based on unreliable methods that are prone to spurious results. See below for a discussion on using the proportion of suicides commited with a gun (PSG) as a proxy for firearm availability when suicide is the outcome.

2 Kleck cites his earlier review (Kleck, 2019), which itself cites his earlier chapter (Kleck, 2018). In this, he uses CDC suicide and suicide attempt data to argue that lethality of firearm and hanging/suffocation suicide attempts are not significantly different because “there is heavy overlap between the confidence intervals surrounding the estimates” (p. 320). However, this is an inappropriate way to make statistical comparisons. Using a Poisson regression with the data in Kleck (2018, Table 17.3), I compared lethality of firearm (258,386 deaths, 308,623 attempts, 83.7% lethal) and hanging/suffocation suicide attempts (118,072 deaths, 154,013 attempts, 76.7% lethal). Firearm suicide attempts are 8.8% more deadly, or conversely, hanging/suffocation attempts are 36.0% more surivavable (both p <.001).

3 “RMPE = recommended minimum effect size representing a ‘practically’ significant effect for social science data” (Ferguson, 2009, p. 533).

Additional information

Notes on contributors

Tyler J. Lane

Tyler J. Lane, DPhil, School of Public Health and Preventive Medicine, Monash University, Melbourne, Australia; Victorian Cancer Registry, Cancer Council Victoria, Melbourne, Australia